This is a maintenance fork. The upstream project's last commit was in March 2026, and it had stopped working against current pi. This fork updates it for compatibility with the current pi version, and folds in three open upstream pull requests that hadn't been merged:
- #25 — don't resume a Claude CLI session that was never created (fixes empty replies when switching to this provider mid-conversation), and surface CLI errors instead of silently returning nothing.
- #26 — fix a false "not authenticated" warning on Claude Code 2.x, and correct the outdated login instructions.
- #29 — let all models use the full thinking-effort range (up to
max), not just Opus.Together these resolve the widely-reported problem where prompting a
pi-claude-climodel just returned an empty response (#3). Credit for the three fixes goes to their original PR authors.
A pi extension that routes LLM calls through the Claude Code CLI as a subprocess. Use your Claude Pro/Max subscription as the LLM backend — no API key, no separate billing.
The extension registers as a custom pi provider exposing all Claude models. Each request spawns a claude -p subprocess using the stream-json wire protocol, with --resume on follow-up turns to reuse the CLI's session state instead of replaying full history. Claude proposes tool calls, pi executes them natively. Custom pi tools are exposed to Claude via a schema-only MCP server.
- Claude Code CLI installed and authenticated (
claudeon PATH) - A Claude Pro or Max subscription
- pi or GSD
Add to ~/.gsd/agent/settings.json:
{
"packages": ["npm:pi-claude-cli"]
}Then select a Claude model via /model in the interactive UI. All Claude models appear under the pi-claude-cli provider.
- Streams text, thinking, and tool call tokens in real-time
- Maps tool names and arguments bidirectionally between Claude and pi
- Exposes custom pi tools to Claude via MCP (schema-only, no execution)
- Break-early pattern prevents Claude CLI from auto-executing tools
- Session resume via
--resumeeliminates history replay on follow-up turns - Configurable thinking effort across the full ladder (low to max) for all models, with elevated mapping for Opus
- Cross-platform subprocess management (Windows, macOS, Linux)
- Inactivity timeout and process registry for cleanup
MIT